A Look Inside Buffett's Battery Bet

[Guest kumar]

 

http://online.wsj.com/article/SB10001424052748704240504574587001841865032.html?mod=googlenews_wsj

 

.....BYD reverted to manual assembly lines with inexpensive labor. The result was cheaply produced batteries of inconsistent quality. Although the company can rightly be proud that it has never faced a battery recall, it scraps 15-30% of its batteries because they fail to meet quality standards—far above the industry average of under 5%....enormous problem for electric-car batteries, in which hundreds of cells, each the size of a pack of cigarettes, must charge and discharge with exact precision. A car battery works only as well as its worst cell. It is far from clear that BYD's labor-intensive process can achieve the uniformity of quality required for electric car batteries.

 

....The all-electric e6 was supposed to reach the U.S. this year, a launch date subsequently pushed out to 2011. But the e6 still hasn't passed Chinese safety tests, let alone more stringent U.S. tests. Advance reviews of prototypes have been scathing: Car and Driver reported, "We drive faster in our driveways."

 

.....MidAmerican's subsidiary in Oregon, PacifiCorp, recently erected a building the size of ten 40-foot storage containers that houses BYD batteries. Those batteries are being tested for the storage of wind-generated energy. BYD's real contribution to Mr. Buffett's portfolio will probably be low-cost, relatively low-tech batteries that store wind and solar power for use on days that are breezeless and cloudy.....

[Dynamic]

I used to work as an R&D Engineer in the leading company in a certain niche of component production in the optical data communications field, with market share over 50% in both leading edge (low-volume, expensive) and mature (high-volume, cheap) products and healthy margins in both. The products contained photodiodes (high yield, not a problem) and semiconductor lasers (yield hits from wafer testing where they're cheap are high, then on subassemblies yields may still be in the 70-85% range for some products). Final testing to ensure compliance at start of life with margin to remain spec-compliant at end-of-life would scrape off a few percent too.

 

We had one line of products we sold in the millions over a few years and had designed almost everything to work with automated manufacture. It was technically difficult to perfect and achieve high yields and actually prevented our customers from second-sourcing because our competitors couldn't do it, especially on their lower volumes and perhaps subject to our patents as obstacles, so our advantage was a double-edged sword that dissuaded customers from wholeheartedly adopting that platform.

 

A similar product range was made using a lot more human input as that was the most cost-efficient approach with our far-east manufacturing operation. Choosing where to take the optional yield hits was an optimization process, i.e. a tight spec and high yield hit at chip level (low cost per item, but more sacrificed) or sub-assembly level (medium cost per item, but fewer sacrificed) or a finished product level (high cost per item, but fewer still sacrificed).

 

Naturally, we took a large hit at chip level, a medium hit at optical subassembly and a small hit at product final test. Sometimes we'd choose a more expensive component in the subassembly that would yield, say, a higher proportion of devices meeting the required power into the optical fiber. Sometimes we'd use a cheaper component if a known proportion of subassemblies would have to be rejected on a laser measurement that could only be made at sub-assembly level, even though a few more would be rejected for insufficient launched power.

 

In the very cheap, very short-range data comms arena, automation was higher, target specs were broader and yields and volumes were higher. Some of the automated manufacture techniques we discontinued were used by our colleagues in this arena and some were used for later product in our group.

 

Conversely in the very expensive long-haul telecommunications arena, automation was lower, yield was far lower, so was volume, and human input was often greater. Additionally, specs were so tight that lasers had to be temperature controlled, contributing to their greater unit cost.

 

It's not necessarily vital to be fully automated or to have the very highest yields to be the leader in terms of volume, costs, profit and quality of delivered product. It may also allow greater flexibility in adapting to demand if a smaller amount of complex capital equipment is required and human resources are readily available. You may also choose to set more demanding limits on measured parameters to afford a higher standard of quality control and prevent reputational loss. In technical products it can also be the right thing to delay launch of a new product by some months to iron out yield and uniformity rather than take liberties with quality control parameters just to start getting the products out of the door (and have to deal with recalls later).

 

After all the the cost isn't proportional to the percentage you reject, it's inversely proportional to the percentage you keep, so 95% yield means unit cost x 1.053, 85% yield = unit cost x 1.176, 70% yield = unit cost x 1.429, not such a high ratio as you might guess by comparing 5% reject rate to 15% or 30%, a factor of 3:1 to 6:1, rather than 1.12:1 to 1.36:1). As products mature over a few years, yield ought to rise, as would production volumes, and selling price is likely to decline. Sometimes, the mix of automation and human input can be adjusted throughout the life cycle of the products to optimize unit cost and volume. Any company can be the leader on one metric, but overall yielded cost is among the most important, not yield by itself. In fact, if yielded cost is already lower on poorer yield, any improvement in yield will create significant further savings.

 

In summary, neither technical leadership nor low cost manufacturing are enough on their own. You have to have sufficient technology to develop acceptable products (which will NOT be exotic and enormously expensive batteries in BYD's case, unless there's definite scope to make them cheap by economies of scale) and then be the low-cost provider of all those. If you can be the volume provider, you get to spread the innovation costs over more products and can hope to make significant advances in technology leadership.

 

The very highest yields are not essential to be the leader in final cost and volume, if the product is significantly cheaper and quality is maintained by product testing screens and a constant focus on yield improvement in the early years of production.

 

Another lesson from my experience - you can be the leader in your field on all these aspects, but you're not immune to the hangover after your customers' binges, or indeed their customers' binges. My job was made redundant a few years after the early 2000's telecom crash when it was clear it was more than just an inventory correction. Reduction in demand was permanent, the boom in investment around Y2K was excessive and there was excess infrastructure and lots of unlit fiber that would take some time to absorb through increased internet usage.